linux/fs/jbd/transaction.c

2224 lines
65 KiB
C
Raw Normal View History

/*
* linux/fs/jbd/transaction.c
*
* Written by Stephen C. Tweedie <sct@redhat.com>, 1998
*
* Copyright 1998 Red Hat corp --- All Rights Reserved
*
* This file is part of the Linux kernel and is made available under
* the terms of the GNU General Public License, version 2, or at your
* option, any later version, incorporated herein by reference.
*
* Generic filesystem transaction handling code; part of the ext2fs
* journaling system.
*
* This file manages transactions (compound commits managed by the
* journaling code) and handles (individual atomic operations by the
* filesystem).
*/
#include <linux/time.h>
#include <linux/fs.h>
#include <linux/jbd.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <linux/timer.h>
#include <linux/mm.h>
#include <linux/highmem.h>
jbd: improve fsync batching There is a flaw with the way jbd handles fsync batching. If we fsync() a file and we were not the last person to run fsync() on this fs then we automatically sleep for 1 jiffie in order to wait for new writers to join into the transaction before forcing the commit. The problem with this is that with really fast storage (ie a Clariion) the time it takes to commit a transaction to disk is way faster than 1 jiffie in most cases, so sleeping means waiting longer with nothing to do than if we just committed the transaction and kept going. Ric Wheeler noticed this when using fs_mark with more than 1 thread, the throughput would plummet as he added more threads. This patch attempts to fix this problem by recording the average time in nanoseconds that it takes to commit a transaction to disk, and what time we started the transaction. If we run an fsync() and we have been running for less time than it takes to commit the transaction to disk, we sleep for the delta amount of time and then commit to disk. We acheive sub-jiffie sleeping using schedule_hrtimeout. This means that the wait time is auto-tuned to the speed of the underlying disk, instead of having this static timeout. I weighted the average according to somebody's comments (Andreas Dilger I think) in order to help normalize random outliers where we take way longer or way less time to commit than the average. I also have a min() check in there to make sure we don't sleep longer than a jiffie in case our storage is super slow, this was requested by Andrew. I unfortunately do not have access to a Clariion, so I had to use a ramdisk to represent a super fast array. I tested with a SATA drive with barrier=1 to make sure there was no regression with local disks, I tested with a 4 way multipathed Apple Xserve RAID array and of course the ramdisk. I ran the following command fs_mark -d /mnt/ext3-test -s 4096 -n 2000 -D 64 -t $i where $i was 2, 4, 8, 16 and 32. I mkfs'ed the fs each time. Here are my results type threads with patch without patch sata 2 24.6 26.3 sata 4 49.2 48.1 sata 8 70.1 67.0 sata 16 104.0 94.1 sata 32 153.6 142.7 xserve 2 246.4 222.0 xserve 4 480.0 440.8 xserve 8 829.5 730.8 xserve 16 1172.7 1026.9 xserve 32 1816.3 1650.5 ramdisk 2 2538.3 1745.6 ramdisk 4 2942.3 661.9 ramdisk 8 2882.5 999.8 ramdisk 16 2738.7 1801.9 ramdisk 32 2541.9 2394.0 Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: Andreas Dilger <adilger@sun.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ric Wheeler <rwheeler@redhat.com> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 02:07:24 +00:00
#include <linux/hrtimer.h>
static void __journal_temp_unlink_buffer(struct journal_head *jh);
/*
* get_transaction: obtain a new transaction_t object.
*
* Simply allocate and initialise a new transaction. Create it in
* RUNNING state and add it to the current journal (which should not
* have an existing running transaction: we only make a new transaction
* once we have started to commit the old one).
*
* Preconditions:
* The journal MUST be locked. We don't perform atomic mallocs on the
* new transaction and we can't block without protecting against other
* processes trying to touch the journal while it is in transition.
*
* Called under j_state_lock
*/
static transaction_t *
get_transaction(journal_t *journal, transaction_t *transaction)
{
transaction->t_journal = journal;
transaction->t_state = T_RUNNING;
jbd: improve fsync batching There is a flaw with the way jbd handles fsync batching. If we fsync() a file and we were not the last person to run fsync() on this fs then we automatically sleep for 1 jiffie in order to wait for new writers to join into the transaction before forcing the commit. The problem with this is that with really fast storage (ie a Clariion) the time it takes to commit a transaction to disk is way faster than 1 jiffie in most cases, so sleeping means waiting longer with nothing to do than if we just committed the transaction and kept going. Ric Wheeler noticed this when using fs_mark with more than 1 thread, the throughput would plummet as he added more threads. This patch attempts to fix this problem by recording the average time in nanoseconds that it takes to commit a transaction to disk, and what time we started the transaction. If we run an fsync() and we have been running for less time than it takes to commit the transaction to disk, we sleep for the delta amount of time and then commit to disk. We acheive sub-jiffie sleeping using schedule_hrtimeout. This means that the wait time is auto-tuned to the speed of the underlying disk, instead of having this static timeout. I weighted the average according to somebody's comments (Andreas Dilger I think) in order to help normalize random outliers where we take way longer or way less time to commit than the average. I also have a min() check in there to make sure we don't sleep longer than a jiffie in case our storage is super slow, this was requested by Andrew. I unfortunately do not have access to a Clariion, so I had to use a ramdisk to represent a super fast array. I tested with a SATA drive with barrier=1 to make sure there was no regression with local disks, I tested with a 4 way multipathed Apple Xserve RAID array and of course the ramdisk. I ran the following command fs_mark -d /mnt/ext3-test -s 4096 -n 2000 -D 64 -t $i where $i was 2, 4, 8, 16 and 32. I mkfs'ed the fs each time. Here are my results type threads with patch without patch sata 2 24.6 26.3 sata 4 49.2 48.1 sata 8 70.1 67.0 sata 16 104.0 94.1 sata 32 153.6 142.7 xserve 2 246.4 222.0 xserve 4 480.0 440.8 xserve 8 829.5 730.8 xserve 16 1172.7 1026.9 xserve 32 1816.3 1650.5 ramdisk 2 2538.3 1745.6 ramdisk 4 2942.3 661.9 ramdisk 8 2882.5 999.8 ramdisk 16 2738.7 1801.9 ramdisk 32 2541.9 2394.0 Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: Andreas Dilger <adilger@sun.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ric Wheeler <rwheeler@redhat.com> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 02:07:24 +00:00
transaction->t_start_time = ktime_get();
transaction->t_tid = journal->j_transaction_sequence++;
transaction->t_expires = jiffies + journal->j_commit_interval;
spin_lock_init(&transaction->t_handle_lock);
/* Set up the commit timer for the new transaction. */
journal->j_commit_timer.expires = round_jiffies(transaction->t_expires);
add_timer(&journal->j_commit_timer);
J_ASSERT(journal->j_running_transaction == NULL);
journal->j_running_transaction = transaction;
return transaction;
}
/*
* Handle management.
*
* A handle_t is an object which represents a single atomic update to a
* filesystem, and which tracks all of the modifications which form part
* of that one update.
*/
/*
* start_this_handle: Given a handle, deal with any locking or stalling
* needed to make sure that there is enough journal space for the handle
* to begin. Attach the handle to a transaction and set up the
* transaction's buffer credits.
*/
static int start_this_handle(journal_t *journal, handle_t *handle)
{
transaction_t *transaction;
int needed;
int nblocks = handle->h_buffer_credits;
transaction_t *new_transaction = NULL;
int ret = 0;
if (nblocks > journal->j_max_transaction_buffers) {
printk(KERN_ERR "JBD: %s wants too many credits (%d > %d)\n",
current->comm, nblocks,
journal->j_max_transaction_buffers);
ret = -ENOSPC;
goto out;
}
alloc_transaction:
if (!journal->j_running_transaction) {
new_transaction = kzalloc(sizeof(*new_transaction),
GFP_NOFS|__GFP_NOFAIL);
if (!new_transaction) {
ret = -ENOMEM;
goto out;
}
}
jbd_debug(3, "New handle %p going live.\n", handle);
repeat:
/*
* We need to hold j_state_lock until t_updates has been incremented,
* for proper journal barrier handling
*/
spin_lock(&journal->j_state_lock);
repeat_locked:
if (is_journal_aborted(journal) ||
(journal->j_errno != 0 && !(journal->j_flags & JFS_ACK_ERR))) {
spin_unlock(&journal->j_state_lock);
ret = -EROFS;
goto out;
}
/* Wait on the journal's transaction barrier if necessary */
if (journal->j_barrier_count) {
spin_unlock(&journal->j_state_lock);
wait_event(journal->j_wait_transaction_locked,
journal->j_barrier_count == 0);
goto repeat;
}
if (!journal->j_running_transaction) {
if (!new_transaction) {
spin_unlock(&journal->j_state_lock);
goto alloc_transaction;
}
get_transaction(journal, new_transaction);
new_transaction = NULL;
}
transaction = journal->j_running_transaction;
/*
* If the current transaction is locked down for commit, wait for the
* lock to be released.
*/
if (transaction->t_state == T_LOCKED) {
DEFINE_WAIT(wait);
prepare_to_wait(&journal->j_wait_transaction_locked,
&wait, TASK_UNINTERRUPTIBLE);
spin_unlock(&journal->j_state_lock);
schedule();
finish_wait(&journal->j_wait_transaction_locked, &wait);
goto repeat;
}
/*
* If there is not enough space left in the log to write all potential
* buffers requested by this operation, we need to stall pending a log
* checkpoint to free some more log space.
*/
spin_lock(&transaction->t_handle_lock);
needed = transaction->t_outstanding_credits + nblocks;
if (needed > journal->j_max_transaction_buffers) {
/*
* If the current transaction is already too large, then start
* to commit it: we can then go back and attach this handle to
* a new transaction.
*/
DEFINE_WAIT(wait);
jbd_debug(2, "Handle %p starting new commit...\n", handle);
spin_unlock(&transaction->t_handle_lock);
prepare_to_wait(&journal->j_wait_transaction_locked, &wait,
TASK_UNINTERRUPTIBLE);
__log_start_commit(journal, transaction->t_tid);
spin_unlock(&journal->j_state_lock);
schedule();
finish_wait(&journal->j_wait_transaction_locked, &wait);
goto repeat;
}
/*
* The commit code assumes that it can get enough log space
* without forcing a checkpoint. This is *critical* for
* correctness: a checkpoint of a buffer which is also
* associated with a committing transaction creates a deadlock,
* so commit simply cannot force through checkpoints.
*
* We must therefore ensure the necessary space in the journal
* *before* starting to dirty potentially checkpointed buffers
* in the new transaction.
*
* The worst part is, any transaction currently committing can
* reduce the free space arbitrarily. Be careful to account for
* those buffers when checkpointing.
*/
/*
* @@@ AKPM: This seems rather over-defensive. We're giving commit
* a _lot_ of headroom: 1/4 of the journal plus the size of
* the committing transaction. Really, we only need to give it
* committing_transaction->t_outstanding_credits plus "enough" for
* the log control blocks.
* Also, this test is inconsitent with the matching one in
* journal_extend().
*/
if (__log_space_left(journal) < jbd_space_needed(journal)) {
jbd_debug(2, "Handle %p waiting for checkpoint...\n", handle);
spin_unlock(&transaction->t_handle_lock);
__log_wait_for_space(journal);
goto repeat_locked;
}
/* OK, account for the buffers that this operation expects to
* use and add the handle to the running transaction. */
handle->h_transaction = transaction;
transaction->t_outstanding_credits += nblocks;
transaction->t_updates++;
transaction->t_handle_count++;
jbd_debug(4, "Handle %p given %d credits (total %d, free %d)\n",
handle, nblocks, transaction->t_outstanding_credits,
__log_space_left(journal));
spin_unlock(&transaction->t_handle_lock);
spin_unlock(&journal->j_state_lock);
out:
if (unlikely(new_transaction)) /* It's usually NULL */
kfree(new_transaction);
return ret;
}
static struct lock_class_key jbd_handle_key;
/* Allocate a new handle. This should probably be in a slab... */
static handle_t *new_handle(int nblocks)
{
handle_t *handle = jbd_alloc_handle(GFP_NOFS);
if (!handle)
return NULL;
memset(handle, 0, sizeof(*handle));
handle->h_buffer_credits = nblocks;
handle->h_ref = 1;
lockdep_init_map(&handle->h_lockdep_map, "jbd_handle", &jbd_handle_key, 0);
return handle;
}
/**
* handle_t *journal_start() - Obtain a new handle.
* @journal: Journal to start transaction on.
* @nblocks: number of block buffer we might modify
*
* We make sure that the transaction can guarantee at least nblocks of
* modified buffers in the log. We block until the log can guarantee
* that much space.
*
* This function is visible to journal users (like ext3fs), so is not
* called with the journal already locked.
*
* Return a pointer to a newly allocated handle, or NULL on failure
*/
handle_t *journal_start(journal_t *journal, int nblocks)
{
handle_t *handle = journal_current_handle();
int err;
if (!journal)
return ERR_PTR(-EROFS);
if (handle) {
J_ASSERT(handle->h_transaction->t_journal == journal);
handle->h_ref++;
return handle;
}
handle = new_handle(nblocks);
if (!handle)
return ERR_PTR(-ENOMEM);
current->journal_info = handle;
err = start_this_handle(journal, handle);
if (err < 0) {
jbd_free_handle(handle);
current->journal_info = NULL;
handle = ERR_PTR(err);
goto out;
}
lock_map_acquire(&handle->h_lockdep_map);
out:
return handle;
}
/**
* int journal_extend() - extend buffer credits.
* @handle: handle to 'extend'
* @nblocks: nr blocks to try to extend by.
*
* Some transactions, such as large extends and truncates, can be done
* atomically all at once or in several stages. The operation requests
* a credit for a number of buffer modications in advance, but can
* extend its credit if it needs more.
*
* journal_extend tries to give the running handle more buffer credits.
* It does not guarantee that allocation - this is a best-effort only.
* The calling process MUST be able to deal cleanly with a failure to
* extend here.
*
* Return 0 on success, non-zero on failure.
*
* return code < 0 implies an error
* return code > 0 implies normal transaction-full status.
*/
int journal_extend(handle_t *handle, int nblocks)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
int result;
int wanted;
result = -EIO;
if (is_handle_aborted(handle))
goto out;
result = 1;
spin_lock(&journal->j_state_lock);
/* Don't extend a locked-down transaction! */
if (handle->h_transaction->t_state != T_RUNNING) {
jbd_debug(3, "denied handle %p %d blocks: "
"transaction not running\n", handle, nblocks);
goto error_out;
}
spin_lock(&transaction->t_handle_lock);
wanted = transaction->t_outstanding_credits + nblocks;
if (wanted > journal->j_max_transaction_buffers) {
jbd_debug(3, "denied handle %p %d blocks: "
"transaction too large\n", handle, nblocks);
goto unlock;
}
if (wanted > __log_space_left(journal)) {
jbd_debug(3, "denied handle %p %d blocks: "
"insufficient log space\n", handle, nblocks);
goto unlock;
}
handle->h_buffer_credits += nblocks;
transaction->t_outstanding_credits += nblocks;
result = 0;
jbd_debug(3, "extended handle %p by %d\n", handle, nblocks);
unlock:
spin_unlock(&transaction->t_handle_lock);
error_out:
spin_unlock(&journal->j_state_lock);
out:
return result;
}
/**
* int journal_restart() - restart a handle.
* @handle: handle to restart
* @nblocks: nr credits requested
*
* Restart a handle for a multi-transaction filesystem
* operation.
*
* If the journal_extend() call above fails to grant new buffer credits
* to a running handle, a call to journal_restart will commit the
* handle's transaction so far and reattach the handle to a new
* transaction capabable of guaranteeing the requested number of
* credits.
*/
int journal_restart(handle_t *handle, int nblocks)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
int ret;
/* If we've had an abort of any type, don't even think about
* actually doing the restart! */
if (is_handle_aborted(handle))
return 0;
/*
* First unlink the handle from its current transaction, and start the
* commit on that.
*/
J_ASSERT(transaction->t_updates > 0);
J_ASSERT(journal_current_handle() == handle);
spin_lock(&journal->j_state_lock);
spin_lock(&transaction->t_handle_lock);
transaction->t_outstanding_credits -= handle->h_buffer_credits;
transaction->t_updates--;
if (!transaction->t_updates)
wake_up(&journal->j_wait_updates);
spin_unlock(&transaction->t_handle_lock);
jbd_debug(2, "restarting handle %p\n", handle);
__log_start_commit(journal, transaction->t_tid);
spin_unlock(&journal->j_state_lock);
handle->h_buffer_credits = nblocks;
ret = start_this_handle(journal, handle);
return ret;
}
/**
* void journal_lock_updates () - establish a transaction barrier.
* @journal: Journal to establish a barrier on.
*
* This locks out any further updates from being started, and blocks
* until all existing updates have completed, returning only once the
* journal is in a quiescent state with no updates running.
*
* The journal lock should not be held on entry.
*/
void journal_lock_updates(journal_t *journal)
{
DEFINE_WAIT(wait);
spin_lock(&journal->j_state_lock);
++journal->j_barrier_count;
/* Wait until there are no running updates */
while (1) {
transaction_t *transaction = journal->j_running_transaction;
if (!transaction)
break;
spin_lock(&transaction->t_handle_lock);
if (!transaction->t_updates) {
spin_unlock(&transaction->t_handle_lock);
break;
}
prepare_to_wait(&journal->j_wait_updates, &wait,
TASK_UNINTERRUPTIBLE);
spin_unlock(&transaction->t_handle_lock);
spin_unlock(&journal->j_state_lock);
schedule();
finish_wait(&journal->j_wait_updates, &wait);
spin_lock(&journal->j_state_lock);
}
spin_unlock(&journal->j_state_lock);
/*
* We have now established a barrier against other normal updates, but
* we also need to barrier against other journal_lock_updates() calls
* to make sure that we serialise special journal-locked operations
* too.
*/
mutex_lock(&journal->j_barrier);
}
/**
* void journal_unlock_updates (journal_t* journal) - release barrier
* @journal: Journal to release the barrier on.
*
* Release a transaction barrier obtained with journal_lock_updates().
*
* Should be called without the journal lock held.
*/
void journal_unlock_updates (journal_t *journal)
{
J_ASSERT(journal->j_barrier_count != 0);
mutex_unlock(&journal->j_barrier);
spin_lock(&journal->j_state_lock);
--journal->j_barrier_count;
spin_unlock(&journal->j_state_lock);
wake_up(&journal->j_wait_transaction_locked);
}
/*
* Report any unexpected dirty buffers which turn up. Normally those
* indicate an error, but they can occur if the user is running (say)
* tune2fs to modify the live filesystem, so we need the option of
* continuing as gracefully as possible. #
*
* The caller should already hold the journal lock and
* j_list_lock spinlock: most callers will need those anyway
* in order to probe the buffer's journaling state safely.
*/
static void jbd_unexpected_dirty_buffer(struct journal_head *jh)
{
int jlist;
/* If this buffer is one which might reasonably be dirty
* --- ie. data, or not part of this journal --- then
* we're OK to leave it alone, but otherwise we need to
* move the dirty bit to the journal's own internal
* JBDDirty bit. */
jlist = jh->b_jlist;
if (jlist == BJ_Metadata || jlist == BJ_Reserved ||
jlist == BJ_Shadow || jlist == BJ_Forget) {
struct buffer_head *bh = jh2bh(jh);
if (test_clear_buffer_dirty(bh))
set_buffer_jbddirty(bh);
}
}
/*
* If the buffer is already part of the current transaction, then there
* is nothing we need to do. If it is already part of a prior
* transaction which we are still committing to disk, then we need to
* make sure that we do not overwrite the old copy: we do copy-out to
* preserve the copy going to disk. We also account the buffer against
* the handle's metadata buffer credits (unless the buffer is already
* part of the transaction, that is).
*
*/
static int
do_get_write_access(handle_t *handle, struct journal_head *jh,
int force_copy)
{
struct buffer_head *bh;
transaction_t *transaction;
journal_t *journal;
int error;
char *frozen_buffer = NULL;
int need_copy = 0;
if (is_handle_aborted(handle))
return -EROFS;
transaction = handle->h_transaction;
journal = transaction->t_journal;
jbd_debug(5, "buffer_head %p, force_copy %d\n", jh, force_copy);
JBUFFER_TRACE(jh, "entry");
repeat:
bh = jh2bh(jh);
/* @@@ Need to check for errors here at some point. */
lock_buffer(bh);
jbd_lock_bh_state(bh);
/* We now hold the buffer lock so it is safe to query the buffer
* state. Is the buffer dirty?
*
* If so, there are two possibilities. The buffer may be
* non-journaled, and undergoing a quite legitimate writeback.
* Otherwise, it is journaled, and we don't expect dirty buffers
* in that state (the buffers should be marked JBD_Dirty
* instead.) So either the IO is being done under our own
* control and this is a bug, or it's a third party IO such as
* dump(8) (which may leave the buffer scheduled for read ---
* ie. locked but not dirty) or tune2fs (which may actually have
* the buffer dirtied, ugh.) */
if (buffer_dirty(bh)) {
/*
* First question: is this buffer already part of the current
* transaction or the existing committing transaction?
*/
if (jh->b_transaction) {
J_ASSERT_JH(jh,
jh->b_transaction == transaction ||
jh->b_transaction ==
journal->j_committing_transaction);
if (jh->b_next_transaction)
J_ASSERT_JH(jh, jh->b_next_transaction ==
transaction);
}
/*
* In any case we need to clean the dirty flag and we must
* do it under the buffer lock to be sure we don't race
* with running write-out.
*/
JBUFFER_TRACE(jh, "Unexpected dirty buffer");
jbd_unexpected_dirty_buffer(jh);
}
unlock_buffer(bh);
error = -EROFS;
if (is_handle_aborted(handle)) {
jbd_unlock_bh_state(bh);
goto out;
}
error = 0;
/*
* The buffer is already part of this transaction if b_transaction or
* b_next_transaction points to it
*/
if (jh->b_transaction == transaction ||
jh->b_next_transaction == transaction)
goto done;
/*
* this is the first time this transaction is touching this buffer,
* reset the modified flag
*/
jh->b_modified = 0;
/*
* If there is already a copy-out version of this buffer, then we don't
* need to make another one
*/
if (jh->b_frozen_data) {
JBUFFER_TRACE(jh, "has frozen data");
J_ASSERT_JH(jh, jh->b_next_transaction == NULL);
jh->b_next_transaction = transaction;
goto done;
}
/* Is there data here we need to preserve? */
if (jh->b_transaction && jh->b_transaction != transaction) {
JBUFFER_TRACE(jh, "owned by older transaction");
J_ASSERT_JH(jh, jh->b_next_transaction == NULL);
J_ASSERT_JH(jh, jh->b_transaction ==
journal->j_committing_transaction);
/* There is one case we have to be very careful about.
* If the committing transaction is currently writing
* this buffer out to disk and has NOT made a copy-out,
* then we cannot modify the buffer contents at all
* right now. The essence of copy-out is that it is the
* extra copy, not the primary copy, which gets
* journaled. If the primary copy is already going to
* disk then we cannot do copy-out here. */
if (jh->b_jlist == BJ_Shadow) {
DEFINE_WAIT_BIT(wait, &bh->b_state, BH_Unshadow);
wait_queue_head_t *wqh;
wqh = bit_waitqueue(&bh->b_state, BH_Unshadow);
JBUFFER_TRACE(jh, "on shadow: sleep");
jbd_unlock_bh_state(bh);
/* commit wakes up all shadow buffers after IO */
for ( ; ; ) {
prepare_to_wait(wqh, &wait.wait,
TASK_UNINTERRUPTIBLE);
if (jh->b_jlist != BJ_Shadow)
break;
schedule();
}
finish_wait(wqh, &wait.wait);
goto repeat;
}
/* Only do the copy if the currently-owning transaction
* still needs it. If it is on the Forget list, the
* committing transaction is past that stage. The
* buffer had better remain locked during the kmalloc,
* but that should be true --- we hold the journal lock
* still and the buffer is already on the BUF_JOURNAL
* list so won't be flushed.
*
* Subtle point, though: if this is a get_undo_access,
* then we will be relying on the frozen_data to contain
* the new value of the committed_data record after the
* transaction, so we HAVE to force the frozen_data copy
* in that case. */
if (jh->b_jlist != BJ_Forget || force_copy) {
JBUFFER_TRACE(jh, "generate frozen data");
if (!frozen_buffer) {
JBUFFER_TRACE(jh, "allocate memory for buffer");
jbd_unlock_bh_state(bh);
frozen_buffer =
jbd_alloc(jh2bh(jh)->b_size,
GFP_NOFS);
if (!frozen_buffer) {
printk(KERN_EMERG
"%s: OOM for frozen_buffer\n",
__func__);
JBUFFER_TRACE(jh, "oom!");
error = -ENOMEM;
jbd_lock_bh_state(bh);
goto done;
}
goto repeat;
}
jh->b_frozen_data = frozen_buffer;
frozen_buffer = NULL;
need_copy = 1;
}
jh->b_next_transaction = transaction;
}
/*
* Finally, if the buffer is not journaled right now, we need to make
* sure it doesn't get written to disk before the caller actually
* commits the new data
*/
if (!jh->b_transaction) {
JBUFFER_TRACE(jh, "no transaction");
J_ASSERT_JH(jh, !jh->b_next_transaction);
jh->b_transaction = transaction;
JBUFFER_TRACE(jh, "file as BJ_Reserved");
spin_lock(&journal->j_list_lock);
__journal_file_buffer(jh, transaction, BJ_Reserved);
spin_unlock(&journal->j_list_lock);
}
done:
if (need_copy) {
struct page *page;
int offset;
char *source;
J_EXPECT_JH(jh, buffer_uptodate(jh2bh(jh)),
"Possible IO failure.\n");
page = jh2bh(jh)->b_page;
offset = ((unsigned long) jh2bh(jh)->b_data) & ~PAGE_MASK;
source = kmap_atomic(page, KM_USER0);
memcpy(jh->b_frozen_data, source+offset, jh2bh(jh)->b_size);
kunmap_atomic(source, KM_USER0);
}
jbd_unlock_bh_state(bh);
/*
* If we are about to journal a buffer, then any revoke pending on it is
* no longer valid
*/
journal_cancel_revoke(handle, jh);
out:
if (unlikely(frozen_buffer)) /* It's usually NULL */
jbd_free(frozen_buffer, bh->b_size);
JBUFFER_TRACE(jh, "exit");
return error;
}
/**
* int journal_get_write_access() - notify intent to modify a buffer for metadata (not data) update.
* @handle: transaction to add buffer modifications to
* @bh: bh to be used for metadata writes
*
* Returns an error code or 0 on success.
*
* In full data journalling mode the buffer may be of type BJ_AsyncData,
* because we're write()ing a buffer which is also part of a shared mapping.
*/
int journal_get_write_access(handle_t *handle, struct buffer_head *bh)
{
struct journal_head *jh = journal_add_journal_head(bh);
int rc;
/* We do not want to get caught playing with fields which the
* log thread also manipulates. Make sure that the buffer
* completes any outstanding IO before proceeding. */
rc = do_get_write_access(handle, jh, 0);
journal_put_journal_head(jh);
return rc;
}
/*
* When the user wants to journal a newly created buffer_head
* (ie. getblk() returned a new buffer and we are going to populate it
* manually rather than reading off disk), then we need to keep the
* buffer_head locked until it has been completely filled with new
* data. In this case, we should be able to make the assertion that
* the bh is not already part of an existing transaction.
*
* The buffer should already be locked by the caller by this point.
* There is no lock ranking violation: it was a newly created,
* unlocked buffer beforehand. */
/**
* int journal_get_create_access () - notify intent to use newly created bh
* @handle: transaction to new buffer to
* @bh: new buffer.
*
* Call this if you create a new bh.
*/
int journal_get_create_access(handle_t *handle, struct buffer_head *bh)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
struct journal_head *jh = journal_add_journal_head(bh);
int err;
jbd_debug(5, "journal_head %p\n", jh);
err = -EROFS;
if (is_handle_aborted(handle))
goto out;
err = 0;
JBUFFER_TRACE(jh, "entry");
/*
* The buffer may already belong to this transaction due to pre-zeroing
* in the filesystem's new_block code. It may also be on the previous,
* committing transaction's lists, but it HAS to be in Forget state in
* that case: the transaction must have deleted the buffer for it to be
* reused here.
*/
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
J_ASSERT_JH(jh, (jh->b_transaction == transaction ||
jh->b_transaction == NULL ||
(jh->b_transaction == journal->j_committing_transaction &&
jh->b_jlist == BJ_Forget)));
J_ASSERT_JH(jh, jh->b_next_transaction == NULL);
J_ASSERT_JH(jh, buffer_locked(jh2bh(jh)));
if (jh->b_transaction == NULL) {
jh->b_transaction = transaction;
/* first access by this transaction */
jh->b_modified = 0;
JBUFFER_TRACE(jh, "file as BJ_Reserved");
__journal_file_buffer(jh, transaction, BJ_Reserved);
} else if (jh->b_transaction == journal->j_committing_transaction) {
/* first access by this transaction */
jh->b_modified = 0;
JBUFFER_TRACE(jh, "set next transaction");
jh->b_next_transaction = transaction;
}
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
/*
* akpm: I added this. ext3_alloc_branch can pick up new indirect
* blocks which contain freed but then revoked metadata. We need
* to cancel the revoke in case we end up freeing it yet again
* and the reallocating as data - this would cause a second revoke,
* which hits an assertion error.
*/
JBUFFER_TRACE(jh, "cancelling revoke");
journal_cancel_revoke(handle, jh);
journal_put_journal_head(jh);
out:
return err;
}
/**
* int journal_get_undo_access() - Notify intent to modify metadata with non-rewindable consequences
* @handle: transaction
* @bh: buffer to undo
*
* Sometimes there is a need to distinguish between metadata which has
* been committed to disk and that which has not. The ext3fs code uses
* this for freeing and allocating space, we have to make sure that we
* do not reuse freed space until the deallocation has been committed,
* since if we overwrote that space we would make the delete
* un-rewindable in case of a crash.
*
* To deal with that, journal_get_undo_access requests write access to a
* buffer for parts of non-rewindable operations such as delete
* operations on the bitmaps. The journaling code must keep a copy of
* the buffer's contents prior to the undo_access call until such time
* as we know that the buffer has definitely been committed to disk.
*
* We never need to know which transaction the committed data is part
* of, buffers touched here are guaranteed to be dirtied later and so
* will be committed to a new transaction in due course, at which point
* we can discard the old committed data pointer.
*
* Returns error number or 0 on success.
*/
int journal_get_undo_access(handle_t *handle, struct buffer_head *bh)
{
int err;
struct journal_head *jh = journal_add_journal_head(bh);
char *committed_data = NULL;
JBUFFER_TRACE(jh, "entry");
/*
* Do this first --- it can drop the journal lock, so we want to
* make sure that obtaining the committed_data is done
* atomically wrt. completion of any outstanding commits.
*/
err = do_get_write_access(handle, jh, 1);
if (err)
goto out;
repeat:
if (!jh->b_committed_data) {
committed_data = jbd_alloc(jh2bh(jh)->b_size, GFP_NOFS);
if (!committed_data) {
printk(KERN_EMERG "%s: No memory for committed data\n",
__func__);
err = -ENOMEM;
goto out;
}
}
jbd_lock_bh_state(bh);
if (!jh->b_committed_data) {
/* Copy out the current buffer contents into the
* preserved, committed copy. */
JBUFFER_TRACE(jh, "generate b_committed data");
if (!committed_data) {
jbd_unlock_bh_state(bh);
goto repeat;
}
jh->b_committed_data = committed_data;
committed_data = NULL;
memcpy(jh->b_committed_data, bh->b_data, bh->b_size);
}
jbd_unlock_bh_state(bh);
out:
journal_put_journal_head(jh);
if (unlikely(committed_data))
jbd_free(committed_data, bh->b_size);
return err;
}
/**
* int journal_dirty_data() - mark a buffer as containing dirty data to be flushed
* @handle: transaction
* @bh: bufferhead to mark
*
* Description:
* Mark a buffer as containing dirty data which needs to be flushed before
* we can commit the current transaction.
*
* The buffer is placed on the transaction's data list and is marked as
* belonging to the transaction.
*
* Returns error number or 0 on success.
*
* journal_dirty_data() can be called via page_launder->ext3_writepage
* by kswapd.
*/
int journal_dirty_data(handle_t *handle, struct buffer_head *bh)
{
journal_t *journal = handle->h_transaction->t_journal;
int need_brelse = 0;
struct journal_head *jh;
int ret = 0;
if (is_handle_aborted(handle))
return ret;
jh = journal_add_journal_head(bh);
JBUFFER_TRACE(jh, "entry");
/*
* The buffer could *already* be dirty. Writeout can start
* at any time.
*/
jbd_debug(4, "jh: %p, tid:%d\n", jh, handle->h_transaction->t_tid);
/*
* What if the buffer is already part of a running transaction?
*
* There are two cases:
* 1) It is part of the current running transaction. Refile it,
* just in case we have allocated it as metadata, deallocated
* it, then reallocated it as data.
* 2) It is part of the previous, still-committing transaction.
* If all we want to do is to guarantee that the buffer will be
* written to disk before this new transaction commits, then
* being sure that the *previous* transaction has this same
* property is sufficient for us! Just leave it on its old
* transaction.
*
* In case (2), the buffer must not already exist as metadata
* --- that would violate write ordering (a transaction is free
* to write its data at any point, even before the previous
* committing transaction has committed). The caller must
* never, ever allow this to happen: there's nothing we can do
* about it in this layer.
*/
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
/* Now that we have bh_state locked, are we really still mapped? */
if (!buffer_mapped(bh)) {
JBUFFER_TRACE(jh, "unmapped buffer, bailing out");
goto no_journal;
}
if (jh->b_transaction) {
JBUFFER_TRACE(jh, "has transaction");
if (jh->b_transaction != handle->h_transaction) {
JBUFFER_TRACE(jh, "belongs to older transaction");
J_ASSERT_JH(jh, jh->b_transaction ==
journal->j_committing_transaction);
/* @@@ IS THIS TRUE ? */
/*
* Not any more. Scenario: someone does a write()
* in data=journal mode. The buffer's transaction has
* moved into commit. Then someone does another
* write() to the file. We do the frozen data copyout
* and set b_next_transaction to point to j_running_t.
* And while we're in that state, someone does a
* writepage() in an attempt to pageout the same area
* of the file via a shared mapping. At present that
* calls journal_dirty_data(), and we get right here.
* It may be too late to journal the data. Simply
* falling through to the next test will suffice: the
* data will be dirty and wil be checkpointed. The
* ordering comments in the next comment block still
* apply.
*/
//J_ASSERT_JH(jh, jh->b_next_transaction == NULL);
/*
* If we're journalling data, and this buffer was
* subject to a write(), it could be metadata, forget
* or shadow against the committing transaction. Now,
* someone has dirtied the same darn page via a mapping
* and it is being writepage()'d.
* We *could* just steal the page from commit, with some
* fancy locking there. Instead, we just skip it -
* don't tie the page's buffers to the new transaction
* at all.
* Implication: if we crash before the writepage() data
* is written into the filesystem, recovery will replay
* the write() data.
*/
if (jh->b_jlist != BJ_None &&
jh->b_jlist != BJ_SyncData &&
jh->b_jlist != BJ_Locked) {
JBUFFER_TRACE(jh, "Not stealing");
goto no_journal;
}
/*
* This buffer may be undergoing writeout in commit. We
* can't return from here and let the caller dirty it
* again because that can cause the write-out loop in
* commit to never terminate.
*/
if (buffer_dirty(bh)) {
get_bh(bh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
need_brelse = 1;
sync_dirty_buffer(bh);
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
/* Since we dropped the lock... */
if (!buffer_mapped(bh)) {
JBUFFER_TRACE(jh, "buffer got unmapped");
goto no_journal;
}
/* The buffer may become locked again at any
time if it is redirtied */
}
/*
* We cannot remove the buffer with io error from the
* committing transaction, because otherwise it would
* miss the error and the commit would not abort.
*/
if (unlikely(!buffer_uptodate(bh))) {
ret = -EIO;
goto no_journal;
}
if (jh->b_transaction != NULL) {
JBUFFER_TRACE(jh, "unfile from commit");
__journal_temp_unlink_buffer(jh);
/* It still points to the committing
* transaction; move it to this one so
* that the refile assert checks are
* happy. */
jh->b_transaction = handle->h_transaction;
}
/* The buffer will be refiled below */
}
/*
* Special case --- the buffer might actually have been
* allocated and then immediately deallocated in the previous,
* committing transaction, so might still be left on that
* transaction's metadata lists.
*/
if (jh->b_jlist != BJ_SyncData && jh->b_jlist != BJ_Locked) {
JBUFFER_TRACE(jh, "not on correct data list: unfile");
J_ASSERT_JH(jh, jh->b_jlist != BJ_Shadow);
__journal_temp_unlink_buffer(jh);
jh->b_transaction = handle->h_transaction;
JBUFFER_TRACE(jh, "file as data");
__journal_file_buffer(jh, handle->h_transaction,
BJ_SyncData);
}
} else {
JBUFFER_TRACE(jh, "not on a transaction");
__journal_file_buffer(jh, handle->h_transaction, BJ_SyncData);
}
no_journal:
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
if (need_brelse) {
BUFFER_TRACE(bh, "brelse");
__brelse(bh);
}
JBUFFER_TRACE(jh, "exit");
journal_put_journal_head(jh);
return ret;
}
/**
* int journal_dirty_metadata() - mark a buffer as containing dirty metadata
* @handle: transaction to add buffer to.
* @bh: buffer to mark
*
* Mark dirty metadata which needs to be journaled as part of the current
* transaction.
*
* The buffer is placed on the transaction's metadata list and is marked
* as belonging to the transaction.
*
* Returns error number or 0 on success.
*
* Special care needs to be taken if the buffer already belongs to the
* current committing transaction (in which case we should have frozen
* data present for that commit). In that case, we don't relink the
* buffer: that only gets done when the old transaction finally
* completes its commit.
*/
int journal_dirty_metadata(handle_t *handle, struct buffer_head *bh)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
struct journal_head *jh = bh2jh(bh);
jbd_debug(5, "journal_head %p\n", jh);
JBUFFER_TRACE(jh, "entry");
if (is_handle_aborted(handle))
goto out;
jbd_lock_bh_state(bh);
if (jh->b_modified == 0) {
/*
* This buffer's got modified and becoming part
* of the transaction. This needs to be done
* once a transaction -bzzz
*/
jh->b_modified = 1;
J_ASSERT_JH(jh, handle->h_buffer_credits > 0);
handle->h_buffer_credits--;
}
/*
* fastpath, to avoid expensive locking. If this buffer is already
* on the running transaction's metadata list there is nothing to do.
* Nobody can take it off again because there is a handle open.
* I _think_ we're OK here with SMP barriers - a mistaken decision will
* result in this test being false, so we go in and take the locks.
*/
if (jh->b_transaction == transaction && jh->b_jlist == BJ_Metadata) {
JBUFFER_TRACE(jh, "fastpath");
J_ASSERT_JH(jh, jh->b_transaction ==
journal->j_running_transaction);
goto out_unlock_bh;
}
set_buffer_jbddirty(bh);
/*
* Metadata already on the current transaction list doesn't
* need to be filed. Metadata on another transaction's list must
* be committing, and will be refiled once the commit completes:
* leave it alone for now.
*/
if (jh->b_transaction != transaction) {
JBUFFER_TRACE(jh, "already on other transaction");
J_ASSERT_JH(jh, jh->b_transaction ==
journal->j_committing_transaction);
J_ASSERT_JH(jh, jh->b_next_transaction == transaction);
/* And this case is illegal: we can't reuse another
* transaction's data buffer, ever. */
goto out_unlock_bh;
}
/* That test should have eliminated the following case: */
J_ASSERT_JH(jh, jh->b_frozen_data == NULL);
JBUFFER_TRACE(jh, "file as BJ_Metadata");
spin_lock(&journal->j_list_lock);
__journal_file_buffer(jh, handle->h_transaction, BJ_Metadata);
spin_unlock(&journal->j_list_lock);
out_unlock_bh:
jbd_unlock_bh_state(bh);
out:
JBUFFER_TRACE(jh, "exit");
return 0;
}
/*
* journal_release_buffer: undo a get_write_access without any buffer
* updates, if the update decided in the end that it didn't need access.
*
*/
void
journal_release_buffer(handle_t *handle, struct buffer_head *bh)
{
BUFFER_TRACE(bh, "entry");
}
/**
* void journal_forget() - bforget() for potentially-journaled buffers.
* @handle: transaction handle
* @bh: bh to 'forget'
*
* We can only do the bforget if there are no commits pending against the
* buffer. If the buffer is dirty in the current running transaction we
* can safely unlink it.
*
* bh may not be a journalled buffer at all - it may be a non-JBD
* buffer which came off the hashtable. Check for this.
*
* Decrements bh->b_count by one.
*
* Allow this call even if the handle has aborted --- it may be part of
* the caller's cleanup after an abort.
*/
int journal_forget (handle_t *handle, struct buffer_head *bh)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
struct journal_head *jh;
int drop_reserve = 0;
int err = 0;
jbd: fix possible journal overflow issues There are several cases where the running transaction can get buffers added to its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers counter end up larger than its t_outstanding_credits counter. This will cause issues when starting new transactions as while we are logging buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes negative, we will report that we need less space in the journal than we actually need, so transactions will be started even though there may not be enough room for them. In the worst case scenario (which admittedly is almost impossible to reproduce) this will result in the journal running out of space. The fix is to only refile buffers from the committing transaction to the running transactions BJ_Modified list when b_modified is set on that journal, which is the only way to be sure if the running transaction has modified that buffer. This patch also fixes an accounting error in journal_forget, it is possible that we can call journal_forget on a buffer without having modified it, only gotten write access to it, so instead of freeing a credit, we only do so if the buffer was modified. The assert will help catch if this problem occurs. Without these two patches I could hit this assert within minutes of running postmark, with them this issue no longer arises. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: <linux-ext4@vger.kernel.org> Acked-by: Jan Kara <jack@ucw.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:16:12 +00:00
int was_modified = 0;
BUFFER_TRACE(bh, "entry");
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
if (!buffer_jbd(bh))
goto not_jbd;
jh = bh2jh(bh);
/* Critical error: attempting to delete a bitmap buffer, maybe?
* Don't do any jbd operations, and return an error. */
if (!J_EXPECT_JH(jh, !jh->b_committed_data,
"inconsistent data on disk")) {
err = -EIO;
goto not_jbd;
}
jbd: fix possible journal overflow issues There are several cases where the running transaction can get buffers added to its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers counter end up larger than its t_outstanding_credits counter. This will cause issues when starting new transactions as while we are logging buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes negative, we will report that we need less space in the journal than we actually need, so transactions will be started even though there may not be enough room for them. In the worst case scenario (which admittedly is almost impossible to reproduce) this will result in the journal running out of space. The fix is to only refile buffers from the committing transaction to the running transactions BJ_Modified list when b_modified is set on that journal, which is the only way to be sure if the running transaction has modified that buffer. This patch also fixes an accounting error in journal_forget, it is possible that we can call journal_forget on a buffer without having modified it, only gotten write access to it, so instead of freeing a credit, we only do so if the buffer was modified. The assert will help catch if this problem occurs. Without these two patches I could hit this assert within minutes of running postmark, with them this issue no longer arises. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: <linux-ext4@vger.kernel.org> Acked-by: Jan Kara <jack@ucw.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:16:12 +00:00
/* keep track of wether or not this transaction modified us */
was_modified = jh->b_modified;
/*
* The buffer's going from the transaction, we must drop
* all references -bzzz
*/
jh->b_modified = 0;
if (jh->b_transaction == handle->h_transaction) {
J_ASSERT_JH(jh, !jh->b_frozen_data);
/* If we are forgetting a buffer which is already part
* of this transaction, then we can just drop it from
* the transaction immediately. */
clear_buffer_dirty(bh);
clear_buffer_jbddirty(bh);
JBUFFER_TRACE(jh, "belongs to current transaction: unfile");
jbd: fix possible journal overflow issues There are several cases where the running transaction can get buffers added to its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers counter end up larger than its t_outstanding_credits counter. This will cause issues when starting new transactions as while we are logging buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes negative, we will report that we need less space in the journal than we actually need, so transactions will be started even though there may not be enough room for them. In the worst case scenario (which admittedly is almost impossible to reproduce) this will result in the journal running out of space. The fix is to only refile buffers from the committing transaction to the running transactions BJ_Modified list when b_modified is set on that journal, which is the only way to be sure if the running transaction has modified that buffer. This patch also fixes an accounting error in journal_forget, it is possible that we can call journal_forget on a buffer without having modified it, only gotten write access to it, so instead of freeing a credit, we only do so if the buffer was modified. The assert will help catch if this problem occurs. Without these two patches I could hit this assert within minutes of running postmark, with them this issue no longer arises. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: <linux-ext4@vger.kernel.org> Acked-by: Jan Kara <jack@ucw.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:16:12 +00:00
/*
* we only want to drop a reference if this transaction
* modified the buffer
*/
if (was_modified)
drop_reserve = 1;
/*
* We are no longer going to journal this buffer.
* However, the commit of this transaction is still
* important to the buffer: the delete that we are now
* processing might obsolete an old log entry, so by
* committing, we can satisfy the buffer's checkpoint.
*
* So, if we have a checkpoint on the buffer, we should
* now refile the buffer on our BJ_Forget list so that
* we know to remove the checkpoint after we commit.
*/
if (jh->b_cp_transaction) {
__journal_temp_unlink_buffer(jh);
__journal_file_buffer(jh, transaction, BJ_Forget);
} else {
__journal_unfile_buffer(jh);
journal_remove_journal_head(bh);
__brelse(bh);
if (!buffer_jbd(bh)) {
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
__bforget(bh);
goto drop;
}
}
} else if (jh->b_transaction) {
J_ASSERT_JH(jh, (jh->b_transaction ==
journal->j_committing_transaction));
/* However, if the buffer is still owned by a prior
* (committing) transaction, we can't drop it yet... */
JBUFFER_TRACE(jh, "belongs to older transaction");
/* ... but we CAN drop it from the new transaction if we
* have also modified it since the original commit. */
if (jh->b_next_transaction) {
J_ASSERT(jh->b_next_transaction == transaction);
jh->b_next_transaction = NULL;
jbd: fix possible journal overflow issues There are several cases where the running transaction can get buffers added to its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers counter end up larger than its t_outstanding_credits counter. This will cause issues when starting new transactions as while we are logging buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes negative, we will report that we need less space in the journal than we actually need, so transactions will be started even though there may not be enough room for them. In the worst case scenario (which admittedly is almost impossible to reproduce) this will result in the journal running out of space. The fix is to only refile buffers from the committing transaction to the running transactions BJ_Modified list when b_modified is set on that journal, which is the only way to be sure if the running transaction has modified that buffer. This patch also fixes an accounting error in journal_forget, it is possible that we can call journal_forget on a buffer without having modified it, only gotten write access to it, so instead of freeing a credit, we only do so if the buffer was modified. The assert will help catch if this problem occurs. Without these two patches I could hit this assert within minutes of running postmark, with them this issue no longer arises. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: <linux-ext4@vger.kernel.org> Acked-by: Jan Kara <jack@ucw.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:16:12 +00:00
/*
* only drop a reference if this transaction modified
* the buffer
*/
if (was_modified)
drop_reserve = 1;
}
}
not_jbd:
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
__brelse(bh);
drop:
if (drop_reserve) {
/* no need to reserve log space for this block -bzzz */
handle->h_buffer_credits++;
}
return err;
}
/**
* int journal_stop() - complete a transaction
* @handle: tranaction to complete.
*
* All done for a particular handle.
*
* There is not much action needed here. We just return any remaining
* buffer credits to the transaction and remove the handle. The only
* complication is that we need to start a commit operation if the
* filesystem is marked for synchronous update.
*
* journal_stop itself will not usually return an error, but it may
* do so in unusual circumstances. In particular, expect it to
* return -EIO if a journal_abort has been executed since the
* transaction began.
*/
int journal_stop(handle_t *handle)
{
transaction_t *transaction = handle->h_transaction;
journal_t *journal = transaction->t_journal;
jbd: improve fsync batching There is a flaw with the way jbd handles fsync batching. If we fsync() a file and we were not the last person to run fsync() on this fs then we automatically sleep for 1 jiffie in order to wait for new writers to join into the transaction before forcing the commit. The problem with this is that with really fast storage (ie a Clariion) the time it takes to commit a transaction to disk is way faster than 1 jiffie in most cases, so sleeping means waiting longer with nothing to do than if we just committed the transaction and kept going. Ric Wheeler noticed this when using fs_mark with more than 1 thread, the throughput would plummet as he added more threads. This patch attempts to fix this problem by recording the average time in nanoseconds that it takes to commit a transaction to disk, and what time we started the transaction. If we run an fsync() and we have been running for less time than it takes to commit the transaction to disk, we sleep for the delta amount of time and then commit to disk. We acheive sub-jiffie sleeping using schedule_hrtimeout. This means that the wait time is auto-tuned to the speed of the underlying disk, instead of having this static timeout. I weighted the average according to somebody's comments (Andreas Dilger I think) in order to help normalize random outliers where we take way longer or way less time to commit than the average. I also have a min() check in there to make sure we don't sleep longer than a jiffie in case our storage is super slow, this was requested by Andrew. I unfortunately do not have access to a Clariion, so I had to use a ramdisk to represent a super fast array. I tested with a SATA drive with barrier=1 to make sure there was no regression with local disks, I tested with a 4 way multipathed Apple Xserve RAID array and of course the ramdisk. I ran the following command fs_mark -d /mnt/ext3-test -s 4096 -n 2000 -D 64 -t $i where $i was 2, 4, 8, 16 and 32. I mkfs'ed the fs each time. Here are my results type threads with patch without patch sata 2 24.6 26.3 sata 4 49.2 48.1 sata 8 70.1 67.0 sata 16 104.0 94.1 sata 32 153.6 142.7 xserve 2 246.4 222.0 xserve 4 480.0 440.8 xserve 8 829.5 730.8 xserve 16 1172.7 1026.9 xserve 32 1816.3 1650.5 ramdisk 2 2538.3 1745.6 ramdisk 4 2942.3 661.9 ramdisk 8 2882.5 999.8 ramdisk 16 2738.7 1801.9 ramdisk 32 2541.9 2394.0 Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: Andreas Dilger <adilger@sun.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ric Wheeler <rwheeler@redhat.com> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 02:07:24 +00:00
int err;
[PATCH] jbd: fix transaction batching Ben points out that: When writing files out using O_SYNC, jbd's 1 jiffy delay results in a significant drop in throughput as the disk sits idle. The patch below results in a 4-5x performance improvement (from 6.5MB/s to ~24-30MB/s on my IDE test box) when writing out files using O_SYNC. So optimise the batching code by omitting it entirely if the process which is doing a sync write is the same as the one which did the most recent sync write. If that's true, we're unlikely to get any other processes joining the transaction. (Has been in -mm for ages - it took me a long time to get on to performance testing it) Numbers, on write-cache-disabled IDE: /usr/bin/time -p synctest -n 10 -uf -t 1 -p 1 dir-name Unpatched: 40 seconds Patched: 35 seconds Batching disabled: 35 seconds This is the problematic single-process-doing-fsync case. With multiple fsyncing processes the numbers are AFACIT unaltered by the patch. Aside: performance testing and instrumentation shows that the transaction batching almost doesn't help (testing with synctest -n 1 -uf -t 100 -p 10 dir-name on non-writeback-caching IDE). This is because by the time one process is running a synchronous commit, a bunch of other processes already have a transaction handle open, so they're all going to batch into the same transaction anyway. The batching seems to offer maybe 5-10% speedup with this workload, but I'm pretty sure it was more important than that when it was first developed 4-odd years ago... Cc: "Stephen C. Tweedie" <sct@redhat.com> Cc: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-05 07:27:54 +00:00
pid_t pid;
J_ASSERT(journal_current_handle() == handle);
if (is_handle_aborted(handle))
err = -EIO;
else {
J_ASSERT(transaction->t_updates > 0);
err = 0;
}
if (--handle->h_ref > 0) {
jbd_debug(4, "h_ref %d -> %d\n", handle->h_ref + 1,
handle->h_ref);
return err;
}
jbd_debug(4, "Handle %p going down\n", handle);
/*
* Implement synchronous transaction batching. If the handle
* was synchronous, don't force a commit immediately. Let's
* yield and let another thread piggyback onto this transaction.
* Keep doing that while new threads continue to arrive.
* It doesn't cost much - we're about to run a commit and sleep
* on IO anyway. Speeds up many-threaded, many-dir operations
* by 30x or more...
[PATCH] jbd: fix transaction batching Ben points out that: When writing files out using O_SYNC, jbd's 1 jiffy delay results in a significant drop in throughput as the disk sits idle. The patch below results in a 4-5x performance improvement (from 6.5MB/s to ~24-30MB/s on my IDE test box) when writing out files using O_SYNC. So optimise the batching code by omitting it entirely if the process which is doing a sync write is the same as the one which did the most recent sync write. If that's true, we're unlikely to get any other processes joining the transaction. (Has been in -mm for ages - it took me a long time to get on to performance testing it) Numbers, on write-cache-disabled IDE: /usr/bin/time -p synctest -n 10 -uf -t 1 -p 1 dir-name Unpatched: 40 seconds Patched: 35 seconds Batching disabled: 35 seconds This is the problematic single-process-doing-fsync case. With multiple fsyncing processes the numbers are AFACIT unaltered by the patch. Aside: performance testing and instrumentation shows that the transaction batching almost doesn't help (testing with synctest -n 1 -uf -t 100 -p 10 dir-name on non-writeback-caching IDE). This is because by the time one process is running a synchronous commit, a bunch of other processes already have a transaction handle open, so they're all going to batch into the same transaction anyway. The batching seems to offer maybe 5-10% speedup with this workload, but I'm pretty sure it was more important than that when it was first developed 4-odd years ago... Cc: "Stephen C. Tweedie" <sct@redhat.com> Cc: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-05 07:27:54 +00:00
*
jbd: improve fsync batching There is a flaw with the way jbd handles fsync batching. If we fsync() a file and we were not the last person to run fsync() on this fs then we automatically sleep for 1 jiffie in order to wait for new writers to join into the transaction before forcing the commit. The problem with this is that with really fast storage (ie a Clariion) the time it takes to commit a transaction to disk is way faster than 1 jiffie in most cases, so sleeping means waiting longer with nothing to do than if we just committed the transaction and kept going. Ric Wheeler noticed this when using fs_mark with more than 1 thread, the throughput would plummet as he added more threads. This patch attempts to fix this problem by recording the average time in nanoseconds that it takes to commit a transaction to disk, and what time we started the transaction. If we run an fsync() and we have been running for less time than it takes to commit the transaction to disk, we sleep for the delta amount of time and then commit to disk. We acheive sub-jiffie sleeping using schedule_hrtimeout. This means that the wait time is auto-tuned to the speed of the underlying disk, instead of having this static timeout. I weighted the average according to somebody's comments (Andreas Dilger I think) in order to help normalize random outliers where we take way longer or way less time to commit than the average. I also have a min() check in there to make sure we don't sleep longer than a jiffie in case our storage is super slow, this was requested by Andrew. I unfortunately do not have access to a Clariion, so I had to use a ramdisk to represent a super fast array. I tested with a SATA drive with barrier=1 to make sure there was no regression with local disks, I tested with a 4 way multipathed Apple Xserve RAID array and of course the ramdisk. I ran the following command fs_mark -d /mnt/ext3-test -s 4096 -n 2000 -D 64 -t $i where $i was 2, 4, 8, 16 and 32. I mkfs'ed the fs each time. Here are my results type threads with patch without patch sata 2 24.6 26.3 sata 4 49.2 48.1 sata 8 70.1 67.0 sata 16 104.0 94.1 sata 32 153.6 142.7 xserve 2 246.4 222.0 xserve 4 480.0 440.8 xserve 8 829.5 730.8 xserve 16 1172.7 1026.9 xserve 32 1816.3 1650.5 ramdisk 2 2538.3 1745.6 ramdisk 4 2942.3 661.9 ramdisk 8 2882.5 999.8 ramdisk 16 2738.7 1801.9 ramdisk 32 2541.9 2394.0 Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: Andreas Dilger <adilger@sun.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ric Wheeler <rwheeler@redhat.com> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 02:07:24 +00:00
* We try and optimize the sleep time against what the underlying disk
* can do, instead of having a static sleep time. This is usefull for
* the case where our storage is so fast that it is more optimal to go
* ahead and force a flush and wait for the transaction to be committed
* than it is to wait for an arbitrary amount of time for new writers to
* join the transaction. We acheive this by measuring how long it takes
* to commit a transaction, and compare it with how long this
* transaction has been running, and if run time < commit time then we
* sleep for the delta and commit. This greatly helps super fast disks
* that would see slowdowns as more threads started doing fsyncs.
*
[PATCH] jbd: fix transaction batching Ben points out that: When writing files out using O_SYNC, jbd's 1 jiffy delay results in a significant drop in throughput as the disk sits idle. The patch below results in a 4-5x performance improvement (from 6.5MB/s to ~24-30MB/s on my IDE test box) when writing out files using O_SYNC. So optimise the batching code by omitting it entirely if the process which is doing a sync write is the same as the one which did the most recent sync write. If that's true, we're unlikely to get any other processes joining the transaction. (Has been in -mm for ages - it took me a long time to get on to performance testing it) Numbers, on write-cache-disabled IDE: /usr/bin/time -p synctest -n 10 -uf -t 1 -p 1 dir-name Unpatched: 40 seconds Patched: 35 seconds Batching disabled: 35 seconds This is the problematic single-process-doing-fsync case. With multiple fsyncing processes the numbers are AFACIT unaltered by the patch. Aside: performance testing and instrumentation shows that the transaction batching almost doesn't help (testing with synctest -n 1 -uf -t 100 -p 10 dir-name on non-writeback-caching IDE). This is because by the time one process is running a synchronous commit, a bunch of other processes already have a transaction handle open, so they're all going to batch into the same transaction anyway. The batching seems to offer maybe 5-10% speedup with this workload, but I'm pretty sure it was more important than that when it was first developed 4-odd years ago... Cc: "Stephen C. Tweedie" <sct@redhat.com> Cc: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-05 07:27:54 +00:00
* But don't do this if this process was the most recent one to
* perform a synchronous write. We do this to detect the case where a
* single process is doing a stream of sync writes. No point in waiting
* for joiners in that case.
*/
[PATCH] jbd: fix transaction batching Ben points out that: When writing files out using O_SYNC, jbd's 1 jiffy delay results in a significant drop in throughput as the disk sits idle. The patch below results in a 4-5x performance improvement (from 6.5MB/s to ~24-30MB/s on my IDE test box) when writing out files using O_SYNC. So optimise the batching code by omitting it entirely if the process which is doing a sync write is the same as the one which did the most recent sync write. If that's true, we're unlikely to get any other processes joining the transaction. (Has been in -mm for ages - it took me a long time to get on to performance testing it) Numbers, on write-cache-disabled IDE: /usr/bin/time -p synctest -n 10 -uf -t 1 -p 1 dir-name Unpatched: 40 seconds Patched: 35 seconds Batching disabled: 35 seconds This is the problematic single-process-doing-fsync case. With multiple fsyncing processes the numbers are AFACIT unaltered by the patch. Aside: performance testing and instrumentation shows that the transaction batching almost doesn't help (testing with synctest -n 1 -uf -t 100 -p 10 dir-name on non-writeback-caching IDE). This is because by the time one process is running a synchronous commit, a bunch of other processes already have a transaction handle open, so they're all going to batch into the same transaction anyway. The batching seems to offer maybe 5-10% speedup with this workload, but I'm pretty sure it was more important than that when it was first developed 4-odd years ago... Cc: "Stephen C. Tweedie" <sct@redhat.com> Cc: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-05 07:27:54 +00:00
pid = current->pid;
if (handle->h_sync && journal->j_last_sync_writer != pid) {
jbd: improve fsync batching There is a flaw with the way jbd handles fsync batching. If we fsync() a file and we were not the last person to run fsync() on this fs then we automatically sleep for 1 jiffie in order to wait for new writers to join into the transaction before forcing the commit. The problem with this is that with really fast storage (ie a Clariion) the time it takes to commit a transaction to disk is way faster than 1 jiffie in most cases, so sleeping means waiting longer with nothing to do than if we just committed the transaction and kept going. Ric Wheeler noticed this when using fs_mark with more than 1 thread, the throughput would plummet as he added more threads. This patch attempts to fix this problem by recording the average time in nanoseconds that it takes to commit a transaction to disk, and what time we started the transaction. If we run an fsync() and we have been running for less time than it takes to commit the transaction to disk, we sleep for the delta amount of time and then commit to disk. We acheive sub-jiffie sleeping using schedule_hrtimeout. This means that the wait time is auto-tuned to the speed of the underlying disk, instead of having this static timeout. I weighted the average according to somebody's comments (Andreas Dilger I think) in order to help normalize random outliers where we take way longer or way less time to commit than the average. I also have a min() check in there to make sure we don't sleep longer than a jiffie in case our storage is super slow, this was requested by Andrew. I unfortunately do not have access to a Clariion, so I had to use a ramdisk to represent a super fast array. I tested with a SATA drive with barrier=1 to make sure there was no regression with local disks, I tested with a 4 way multipathed Apple Xserve RAID array and of course the ramdisk. I ran the following command fs_mark -d /mnt/ext3-test -s 4096 -n 2000 -D 64 -t $i where $i was 2, 4, 8, 16 and 32. I mkfs'ed the fs each time. Here are my results type threads with patch without patch sata 2 24.6 26.3 sata 4 49.2 48.1 sata 8 70.1 67.0 sata 16 104.0 94.1 sata 32 153.6 142.7 xserve 2 246.4 222.0 xserve 4 480.0 440.8 xserve 8 829.5 730.8 xserve 16 1172.7 1026.9 xserve 32 1816.3 1650.5 ramdisk 2 2538.3 1745.6 ramdisk 4 2942.3 661.9 ramdisk 8 2882.5 999.8 ramdisk 16 2738.7 1801.9 ramdisk 32 2541.9 2394.0 Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: Andreas Dilger <adilger@sun.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ric Wheeler <rwheeler@redhat.com> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 02:07:24 +00:00
u64 commit_time, trans_time;
[PATCH] jbd: fix transaction batching Ben points out that: When writing files out using O_SYNC, jbd's 1 jiffy delay results in a significant drop in throughput as the disk sits idle. The patch below results in a 4-5x performance improvement (from 6.5MB/s to ~24-30MB/s on my IDE test box) when writing out files using O_SYNC. So optimise the batching code by omitting it entirely if the process which is doing a sync write is the same as the one which did the most recent sync write. If that's true, we're unlikely to get any other processes joining the transaction. (Has been in -mm for ages - it took me a long time to get on to performance testing it) Numbers, on write-cache-disabled IDE: /usr/bin/time -p synctest -n 10 -uf -t 1 -p 1 dir-name Unpatched: 40 seconds Patched: 35 seconds Batching disabled: 35 seconds This is the problematic single-process-doing-fsync case. With multiple fsyncing processes the numbers are AFACIT unaltered by the patch. Aside: performance testing and instrumentation shows that the transaction batching almost doesn't help (testing with synctest -n 1 -uf -t 100 -p 10 dir-name on non-writeback-caching IDE). This is because by the time one process is running a synchronous commit, a bunch of other processes already have a transaction handle open, so they're all going to batch into the same transaction anyway. The batching seems to offer maybe 5-10% speedup with this workload, but I'm pretty sure it was more important than that when it was first developed 4-odd years ago... Cc: "Stephen C. Tweedie" <sct@redhat.com> Cc: Benjamin LaHaise <bcrl@kvack.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-05 07:27:54 +00:00
journal->j_last_sync_writer = pid;
jbd: improve fsync batching There is a flaw with the way jbd handles fsync batching. If we fsync() a file and we were not the last person to run fsync() on this fs then we automatically sleep for 1 jiffie in order to wait for new writers to join into the transaction before forcing the commit. The problem with this is that with really fast storage (ie a Clariion) the time it takes to commit a transaction to disk is way faster than 1 jiffie in most cases, so sleeping means waiting longer with nothing to do than if we just committed the transaction and kept going. Ric Wheeler noticed this when using fs_mark with more than 1 thread, the throughput would plummet as he added more threads. This patch attempts to fix this problem by recording the average time in nanoseconds that it takes to commit a transaction to disk, and what time we started the transaction. If we run an fsync() and we have been running for less time than it takes to commit the transaction to disk, we sleep for the delta amount of time and then commit to disk. We acheive sub-jiffie sleeping using schedule_hrtimeout. This means that the wait time is auto-tuned to the speed of the underlying disk, instead of having this static timeout. I weighted the average according to somebody's comments (Andreas Dilger I think) in order to help normalize random outliers where we take way longer or way less time to commit than the average. I also have a min() check in there to make sure we don't sleep longer than a jiffie in case our storage is super slow, this was requested by Andrew. I unfortunately do not have access to a Clariion, so I had to use a ramdisk to represent a super fast array. I tested with a SATA drive with barrier=1 to make sure there was no regression with local disks, I tested with a 4 way multipathed Apple Xserve RAID array and of course the ramdisk. I ran the following command fs_mark -d /mnt/ext3-test -s 4096 -n 2000 -D 64 -t $i where $i was 2, 4, 8, 16 and 32. I mkfs'ed the fs each time. Here are my results type threads with patch without patch sata 2 24.6 26.3 sata 4 49.2 48.1 sata 8 70.1 67.0 sata 16 104.0 94.1 sata 32 153.6 142.7 xserve 2 246.4 222.0 xserve 4 480.0 440.8 xserve 8 829.5 730.8 xserve 16 1172.7 1026.9 xserve 32 1816.3 1650.5 ramdisk 2 2538.3 1745.6 ramdisk 4 2942.3 661.9 ramdisk 8 2882.5 999.8 ramdisk 16 2738.7 1801.9 ramdisk 32 2541.9 2394.0 Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: Andreas Dilger <adilger@sun.com> Cc: Arjan van de Ven <arjan@infradead.org> Cc: Ric Wheeler <rwheeler@redhat.com> Cc: <linux-ext4@vger.kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2009-01-08 02:07:24 +00:00
spin_lock(&journal->j_state_lock);
commit_time = journal->j_average_commit_time;
spin_unlock(&journal->j_state_lock);
trans_time = ktime_to_ns(ktime_sub(ktime_get(),
transaction->t_start_time));
commit_time = min_t(u64, commit_time,
1000*jiffies_to_usecs(1));
if (trans_time < commit_time) {
ktime_t expires = ktime_add_ns(ktime_get(),
commit_time);
set_current_state(TASK_UNINTERRUPTIBLE);
schedule_hrtimeout(&expires, HRTIMER_MODE_ABS);
}
}
current->journal_info = NULL;
spin_lock(&journal->j_state_lock);
spin_lock(&transaction->t_handle_lock);
transaction->t_outstanding_credits -= handle->h_buffer_credits;
transaction->t_updates--;
if (!transaction->t_updates) {
wake_up(&journal->j_wait_updates);
if (journal->j_barrier_count)
wake_up(&journal->j_wait_transaction_locked);
}
/*
* If the handle is marked SYNC, we need to set another commit
* going! We also want to force a commit if the current
* transaction is occupying too much of the log, or if the
* transaction is too old now.
*/
if (handle->h_sync ||
transaction->t_outstanding_credits >
journal->j_max_transaction_buffers ||
time_after_eq(jiffies, transaction->t_expires)) {
/* Do this even for aborted journals: an abort still
* completes the commit thread, it just doesn't write
* anything to disk. */
tid_t tid = transaction->t_tid;
spin_unlock(&transaction->t_handle_lock);
jbd_debug(2, "transaction too old, requesting commit for "
"handle %p\n", handle);
/* This is non-blocking */
__log_start_commit(journal, transaction->t_tid);
spin_unlock(&journal->j_state_lock);
/*
* Special case: JFS_SYNC synchronous updates require us
* to wait for the commit to complete.
*/
if (handle->h_sync && !(current->flags & PF_MEMALLOC))
err = log_wait_commit(journal, tid);
} else {
spin_unlock(&transaction->t_handle_lock);
spin_unlock(&journal->j_state_lock);
}
lock_map_release(&handle->h_lockdep_map);
jbd_free_handle(handle);
return err;
}
/**
* int journal_force_commit() - force any uncommitted transactions
* @journal: journal to force
*
* For synchronous operations: force any uncommitted transactions
* to disk. May seem kludgy, but it reuses all the handle batching
* code in a very simple manner.
*/
int journal_force_commit(journal_t *journal)
{
handle_t *handle;
int ret;
handle = journal_start(journal, 1);
if (IS_ERR(handle)) {
ret = PTR_ERR(handle);
} else {
handle->h_sync = 1;
ret = journal_stop(handle);
}
return ret;
}
/*
*
* List management code snippets: various functions for manipulating the
* transaction buffer lists.
*
*/
/*
* Append a buffer to a transaction list, given the transaction's list head
* pointer.
*
* j_list_lock is held.
*
* jbd_lock_bh_state(jh2bh(jh)) is held.
*/
static inline void
__blist_add_buffer(struct journal_head **list, struct journal_head *jh)
{
if (!*list) {
jh->b_tnext = jh->b_tprev = jh;
*list = jh;
} else {
/* Insert at the tail of the list to preserve order */
struct journal_head *first = *list, *last = first->b_tprev;
jh->b_tprev = last;
jh->b_tnext = first;
last->b_tnext = first->b_tprev = jh;
}
}
/*
* Remove a buffer from a transaction list, given the transaction's list
* head pointer.
*
* Called with j_list_lock held, and the journal may not be locked.
*
* jbd_lock_bh_state(jh2bh(jh)) is held.
*/
static inline void
__blist_del_buffer(struct journal_head **list, struct journal_head *jh)
{
if (*list == jh) {
*list = jh->b_tnext;
if (*list == jh)
*list = NULL;
}
jh->b_tprev->b_tnext = jh->b_tnext;
jh->b_tnext->b_tprev = jh->b_tprev;
}
/*
* Remove a buffer from the appropriate transaction list.
*
* Note that this function can *change* the value of
* bh->b_transaction->t_sync_datalist, t_buffers, t_forget,
* t_iobuf_list, t_shadow_list, t_log_list or t_reserved_list. If the caller
* is holding onto a copy of one of thee pointers, it could go bad.
* Generally the caller needs to re-read the pointer from the transaction_t.
*
* Called under j_list_lock. The journal may not be locked.
*/
static void __journal_temp_unlink_buffer(struct journal_head *jh)
{
struct journal_head **list = NULL;
transaction_t *transaction;
struct buffer_head *bh = jh2bh(jh);
J_ASSERT_JH(jh, jbd_is_locked_bh_state(bh));
transaction = jh->b_transaction;
if (transaction)
assert_spin_locked(&transaction->t_journal->j_list_lock);
J_ASSERT_JH(jh, jh->b_jlist < BJ_Types);
if (jh->b_jlist != BJ_None)
J_ASSERT_JH(jh, transaction != NULL);
switch (jh->b_jlist) {
case BJ_None:
return;
case BJ_SyncData:
list = &transaction->t_sync_datalist;
break;
case BJ_Metadata:
transaction->t_nr_buffers--;
J_ASSERT_JH(jh, transaction->t_nr_buffers >= 0);
list = &transaction->t_buffers;
break;
case BJ_Forget:
list = &transaction->t_forget;
break;
case BJ_IO:
list = &transaction->t_iobuf_list;
break;
case BJ_Shadow:
list = &transaction->t_shadow_list;
break;
case BJ_LogCtl:
list = &transaction->t_log_list;
break;
case BJ_Reserved:
list = &transaction->t_reserved_list;
break;
case BJ_Locked:
list = &transaction->t_locked_list;
break;
}
__blist_del_buffer(list, jh);
jh->b_jlist = BJ_None;
if (test_clear_buffer_jbddirty(bh))
mark_buffer_dirty(bh); /* Expose it to the VM */
}
void __journal_unfile_buffer(struct journal_head *jh)
{
__journal_temp_unlink_buffer(jh);
jh->b_transaction = NULL;
}
void journal_unfile_buffer(journal_t *journal, struct journal_head *jh)
{
jbd_lock_bh_state(jh2bh(jh));
spin_lock(&journal->j_list_lock);
__journal_unfile_buffer(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(jh2bh(jh));
}
/*
* Called from journal_try_to_free_buffers().
*
* Called under jbd_lock_bh_state(bh)
*/
static void
__journal_try_to_free_buffer(journal_t *journal, struct buffer_head *bh)
{
struct journal_head *jh;
jh = bh2jh(bh);
if (buffer_locked(bh) || buffer_dirty(bh))
goto out;
if (jh->b_next_transaction != NULL)
goto out;
spin_lock(&journal->j_list_lock);
if (jh->b_transaction != NULL && jh->b_cp_transaction == NULL) {
if (jh->b_jlist == BJ_SyncData || jh->b_jlist == BJ_Locked) {
/* A written-back ordered data buffer */
JBUFFER_TRACE(jh, "release data");
__journal_unfile_buffer(jh);
journal_remove_journal_head(bh);
__brelse(bh);
}
} else if (jh->b_cp_transaction != NULL && jh->b_transaction == NULL) {
/* written-back checkpointed metadata buffer */
if (jh->b_jlist == BJ_None) {
JBUFFER_TRACE(jh, "remove from checkpoint list");
__journal_remove_checkpoint(jh);
journal_remove_journal_head(bh);
__brelse(bh);
}
}
spin_unlock(&journal->j_list_lock);
out:
return;
}
/*
* journal_try_to_free_buffers() could race with journal_commit_transaction()
* The latter might still hold the a count on buffers when inspecting
* them on t_syncdata_list or t_locked_list.
*
* journal_try_to_free_buffers() will call this function to
* wait for the current transaction to finish syncing data buffers, before
* tryinf to free that buffer.
*
* Called with journal->j_state_lock held.
*/
static void journal_wait_for_transaction_sync_data(journal_t *journal)
{
transaction_t *transaction = NULL;
tid_t tid;
spin_lock(&journal->j_state_lock);
transaction = journal->j_committing_transaction;
if (!transaction) {
spin_unlock(&journal->j_state_lock);
return;
}
tid = transaction->t_tid;
spin_unlock(&journal->j_state_lock);
log_wait_commit(journal, tid);
}
/**
* int journal_try_to_free_buffers() - try to free page buffers.
* @journal: journal for operation
* @page: to try and free
* @gfp_mask: we use the mask to detect how hard should we try to release
* buffers. If __GFP_WAIT and __GFP_FS is set, we wait for commit code to
* release the buffers.
*
*
* For all the buffers on this page,
* if they are fully written out ordered data, move them onto BUF_CLEAN
* so try_to_free_buffers() can reap them.
*
* This function returns non-zero if we wish try_to_free_buffers()
* to be called. We do this if the page is releasable by try_to_free_buffers().
* We also do it if the page has locked or dirty buffers and the caller wants
* us to perform sync or async writeout.
*
* This complicates JBD locking somewhat. We aren't protected by the
* BKL here. We wish to remove the buffer from its committing or
* running transaction's ->t_datalist via __journal_unfile_buffer.
*
* This may *change* the value of transaction_t->t_datalist, so anyone
* who looks at t_datalist needs to lock against this function.
*
* Even worse, someone may be doing a journal_dirty_data on this
* buffer. So we need to lock against that. journal_dirty_data()
* will come out of the lock with the buffer dirty, which makes it
* ineligible for release here.
*
* Who else is affected by this? hmm... Really the only contender
* is do_get_write_access() - it could be looking at the buffer while
* journal_try_to_free_buffer() is changing its state. But that
* cannot happen because we never reallocate freed data as metadata
* while the data is part of a transaction. Yes?
*
* Return 0 on failure, 1 on success
*/
int journal_try_to_free_buffers(journal_t *journal,
struct page *page, gfp_t gfp_mask)
{
struct buffer_head *head;
struct buffer_head *bh;
int ret = 0;
J_ASSERT(PageLocked(page));
head = page_buffers(page);
bh = head;
do {
struct journal_head *jh;
/*
* We take our own ref against the journal_head here to avoid
* having to add tons of locking around each instance of
* journal_remove_journal_head() and journal_put_journal_head().
*/
jh = journal_grab_journal_head(bh);
if (!jh)
continue;
jbd_lock_bh_state(bh);
__journal_try_to_free_buffer(journal, bh);
journal_put_journal_head(jh);
jbd_unlock_bh_state(bh);
if (buffer_jbd(bh))
goto busy;
} while ((bh = bh->b_this_page) != head);
ret = try_to_free_buffers(page);
/*
* There are a number of places where journal_try_to_free_buffers()
* could race with journal_commit_transaction(), the later still
* holds the reference to the buffers to free while processing them.
* try_to_free_buffers() failed to free those buffers. Some of the
* caller of releasepage() request page buffers to be dropped, otherwise
* treat the fail-to-free as errors (such as generic_file_direct_IO())
*
* So, if the caller of try_to_release_page() wants the synchronous
* behaviour(i.e make sure buffers are dropped upon return),
* let's wait for the current transaction to finish flush of
* dirty data buffers, then try to free those buffers again,
* with the journal locked.
*/
if (ret == 0 && (gfp_mask & __GFP_WAIT) && (gfp_mask & __GFP_FS)) {
journal_wait_for_transaction_sync_data(journal);
ret = try_to_free_buffers(page);
}
busy:
return ret;
}
/*
* This buffer is no longer needed. If it is on an older transaction's
* checkpoint list we need to record it on this transaction's forget list
* to pin this buffer (and hence its checkpointing transaction) down until
* this transaction commits. If the buffer isn't on a checkpoint list, we
* release it.
* Returns non-zero if JBD no longer has an interest in the buffer.
*
* Called under j_list_lock.
*
* Called under jbd_lock_bh_state(bh).
*/
static int __dispose_buffer(struct journal_head *jh, transaction_t *transaction)
{
int may_free = 1;
struct buffer_head *bh = jh2bh(jh);
__journal_unfile_buffer(jh);
if (jh->b_cp_transaction) {
JBUFFER_TRACE(jh, "on running+cp transaction");
__journal_file_buffer(jh, transaction, BJ_Forget);
clear_buffer_jbddirty(bh);
may_free = 0;
} else {
JBUFFER_TRACE(jh, "on running transaction");
journal_remove_journal_head(bh);
__brelse(bh);
}
return may_free;
}
/*
* journal_invalidatepage
*
* This code is tricky. It has a number of cases to deal with.
*
* There are two invariants which this code relies on:
*
* i_size must be updated on disk before we start calling invalidatepage on the
* data.
*
* This is done in ext3 by defining an ext3_setattr method which
* updates i_size before truncate gets going. By maintaining this
* invariant, we can be sure that it is safe to throw away any buffers
* attached to the current transaction: once the transaction commits,
* we know that the data will not be needed.
*
* Note however that we can *not* throw away data belonging to the
* previous, committing transaction!
*
* Any disk blocks which *are* part of the previous, committing
* transaction (and which therefore cannot be discarded immediately) are
* not going to be reused in the new running transaction
*
* The bitmap committed_data images guarantee this: any block which is
* allocated in one transaction and removed in the next will be marked
* as in-use in the committed_data bitmap, so cannot be reused until
* the next transaction to delete the block commits. This means that
* leaving committing buffers dirty is quite safe: the disk blocks
* cannot be reallocated to a different file and so buffer aliasing is
* not possible.
*
*
* The above applies mainly to ordered data mode. In writeback mode we
* don't make guarantees about the order in which data hits disk --- in
* particular we don't guarantee that new dirty data is flushed before
* transaction commit --- so it is always safe just to discard data
* immediately in that mode. --sct
*/
/*
* The journal_unmap_buffer helper function returns zero if the buffer
* concerned remains pinned as an anonymous buffer belonging to an older
* transaction.
*
* We're outside-transaction here. Either or both of j_running_transaction
* and j_committing_transaction may be NULL.
*/
static int journal_unmap_buffer(journal_t *journal, struct buffer_head *bh)
{
transaction_t *transaction;
struct journal_head *jh;
int may_free = 1;
int ret;
BUFFER_TRACE(bh, "entry");
/*
* It is safe to proceed here without the j_list_lock because the
* buffers cannot be stolen by try_to_free_buffers as long as we are
* holding the page lock. --sct
*/
if (!buffer_jbd(bh))
goto zap_buffer_unlocked;
spin_lock(&journal->j_state_lock);
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
jh = journal_grab_journal_head(bh);
if (!jh)
goto zap_buffer_no_jh;
transaction = jh->b_transaction;
if (transaction == NULL) {
/* First case: not on any transaction. If it
* has no checkpoint link, then we can zap it:
* it's a writeback-mode buffer so we don't care
* if it hits disk safely. */
if (!jh->b_cp_transaction) {
JBUFFER_TRACE(jh, "not on any transaction: zap");
goto zap_buffer;
}
if (!buffer_dirty(bh)) {
/* bdflush has written it. We can drop it now */
goto zap_buffer;
}
/* OK, it must be in the journal but still not
* written fully to disk: it's metadata or
* journaled data... */
if (journal->j_running_transaction) {
/* ... and once the current transaction has
* committed, the buffer won't be needed any
* longer. */
JBUFFER_TRACE(jh, "checkpointed: add to BJ_Forget");
ret = __dispose_buffer(jh,
journal->j_running_transaction);
journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
return ret;
} else {
/* There is no currently-running transaction. So the
* orphan record which we wrote for this file must have
* passed into commit. We must attach this buffer to
* the committing transaction, if it exists. */
if (journal->j_committing_transaction) {
JBUFFER_TRACE(jh, "give to committing trans");
ret = __dispose_buffer(jh,
journal->j_committing_transaction);
journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
return ret;
} else {
/* The orphan record's transaction has
* committed. We can cleanse this buffer */
clear_buffer_jbddirty(bh);
goto zap_buffer;
}
}
} else if (transaction == journal->j_committing_transaction) {
JBUFFER_TRACE(jh, "on committing transaction");
[PATCH] jbd dirty buffer leak fix This fixes the lots-of-fsx-linux-instances-cause-a-slow-leak bug. It's been there since 2.6.6, caused by: ftp://ftp.kernel.org/pub/linux/kernel/people/akpm/patches/2.6/2.6.5/2.6.5-mm4/broken-out/jbd-move-locked-buffers.patch That patch moves under-writeout ordered-data buffers onto a separate journal list during commit. It took out the old code which was based on a single list. The old code (necessarily) had logic which would restart I/O against buffers which had been redirtied while they were on the committing transaction's t_sync_datalist list. The new code only writes buffers once, ignoring redirtyings by a later transaction, which is good. But over on the truncate side of things, in journal_unmap_buffer(), we're treating buffers on the t_locked_list as inviolable things which belong to the committing transaction, and we just leave them alone during concurrent truncate-vs-commit. The net effect is that when truncate tries to invalidate a page whose buffers are on t_locked_list and have been redirtied, journal_unmap_buffer() just leaves those buffers alone. truncate will remove the page from its mapping and we end up with an anonymous clean page with dirty buffers, which is an illegal state for a page. The JBD commit will not clean those buffers as they are removed from t_locked_list. The VM (try_to_free_buffers) cannot reclaim these pages. The patch teaches journal_unmap_buffer() about buffers which are on the committing transaction's t_locked_list. These buffers have been written and I/O has completed. We can take them off the transaction and undirty them within the context of journal_invalidatepage()->journal_unmap_buffer(). Acked-by: "Stephen C. Tweedie" <sct@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-04-16 22:26:36 +00:00
if (jh->b_jlist == BJ_Locked) {
/*
* The buffer is on the committing transaction's locked
* list. We have the buffer locked, so I/O has
* completed. So we can nail the buffer now.
*/
may_free = __dispose_buffer(jh, transaction);
goto zap_buffer;
}
/*
* If it is committing, we simply cannot touch it. We
* can remove it's next_transaction pointer from the
* running transaction if that is set, but nothing
* else. */
set_buffer_freed(bh);
if (jh->b_next_transaction) {
J_ASSERT(jh->b_next_transaction ==
journal->j_running_transaction);
jh->b_next_transaction = NULL;
}
journal_put_journal_head(jh);
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
return 0;
} else {
/* Good, the buffer belongs to the running transaction.
* We are writing our own transaction's data, not any
* previous one's, so it is safe to throw it away
* (remember that we expect the filesystem to have set
* i_size already for this truncate so recovery will not
* expose the disk blocks we are discarding here.) */
J_ASSERT_JH(jh, transaction == journal->j_running_transaction);
JBUFFER_TRACE(jh, "on running transaction");
may_free = __dispose_buffer(jh, transaction);
}
zap_buffer:
journal_put_journal_head(jh);
zap_buffer_no_jh:
spin_unlock(&journal->j_list_lock);
jbd_unlock_bh_state(bh);
spin_unlock(&journal->j_state_lock);
zap_buffer_unlocked:
clear_buffer_dirty(bh);
J_ASSERT_BH(bh, !buffer_jbddirty(bh));
clear_buffer_mapped(bh);
clear_buffer_req(bh);
clear_buffer_new(bh);
bh->b_bdev = NULL;
return may_free;
}
/**
fs: fix kernel-doc notation warnings Fix kernel-doc notation warnings in fs/. Warning(mmotm-2008-0314-1449//fs/super.c:560): missing initial short description on line: * mark_files_ro Warning(mmotm-2008-0314-1449//fs/locks.c:1277): missing initial short description on line: * lease_get_mtime Warning(mmotm-2008-0314-1449//fs/locks.c:1277): missing initial short description on line: * lease_get_mtime Warning(mmotm-2008-0314-1449//fs/namei.c:1368): missing initial short description on line: * lookup_one_len: filesystem helper to lookup single pathname component Warning(mmotm-2008-0314-1449//fs/buffer.c:3221): missing initial short description on line: * bh_uptodate_or_lock: Test whether the buffer is uptodate Warning(mmotm-2008-0314-1449//fs/buffer.c:3240): missing initial short description on line: * bh_submit_read: Submit a locked buffer for reading Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:30): missing initial short description on line: * writeback_acquire: attempt to get exclusive writeback access to a device Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:47): missing initial short description on line: * writeback_in_progress: determine whether there is writeback in progress Warning(mmotm-2008-0314-1449//fs/fs-writeback.c:58): missing initial short description on line: * writeback_release: relinquish exclusive writeback access against a device. Warning(mmotm-2008-0314-1449//include/linux/jbd.h:351): contents before sections Warning(mmotm-2008-0314-1449//include/linux/jbd.h:561): contents before sections Warning(mmotm-2008-0314-1449//fs/jbd/transaction.c:1935): missing initial short description on line: * void journal_invalidatepage() Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-03-20 00:01:00 +00:00
* void journal_invalidatepage() - invalidate a journal page
* @journal: journal to use for flush
* @page: page to flush
* @offset: length of page to invalidate.
*
* Reap page buffers containing data after offset in page.
*/
void journal_invalidatepage(journal_t *journal,
struct page *page,
unsigned long offset)
{
struct buffer_head *head, *bh, *next;
unsigned int curr_off = 0;
int may_free = 1;
if (!PageLocked(page))
BUG();
if (!page_has_buffers(page))
return;
/* We will potentially be playing with lists other than just the
* data lists (especially for journaled data mode), so be
* cautious in our locking. */
head = bh = page_buffers(page);
do {
unsigned int next_off = curr_off + bh->b_size;
next = bh->b_this_page;
if (offset <= curr_off) {
/* This block is wholly outside the truncation point */
lock_buffer(bh);
may_free &= journal_unmap_buffer(journal, bh);
unlock_buffer(bh);
}
curr_off = next_off;
bh = next;
} while (bh != head);
if (!offset) {
if (may_free && try_to_free_buffers(page))
J_ASSERT(!page_has_buffers(page));
}
}
/*
* File a buffer on the given transaction list.
*/
void __journal_file_buffer(struct journal_head *jh,
transaction_t *transaction, int jlist)
{
struct journal_head **list = NULL;
int was_dirty = 0;
struct buffer_head *bh = jh2bh(jh);
J_ASSERT_JH(jh, jbd_is_locked_bh_state(bh));
assert_spin_locked(&transaction->t_journal->j_list_lock);
J_ASSERT_JH(jh, jh->b_jlist < BJ_Types);
J_ASSERT_JH(jh, jh->b_transaction == transaction ||
jh->b_transaction == NULL);
if (jh->b_transaction && jh->b_jlist == jlist)
return;
/* The following list of buffer states needs to be consistent
* with __jbd_unexpected_dirty_buffer()'s handling of dirty
* state. */
if (jlist == BJ_Metadata || jlist == BJ_Reserved ||
jlist == BJ_Shadow || jlist == BJ_Forget) {
if (test_clear_buffer_dirty(bh) ||
test_clear_buffer_jbddirty(bh))
was_dirty = 1;
}
if (jh->b_transaction)
__journal_temp_unlink_buffer(jh);
jh->b_transaction = transaction;
switch (jlist) {
case BJ_None:
J_ASSERT_JH(jh, !jh->b_committed_data);
J_ASSERT_JH(jh, !jh->b_frozen_data);
return;
case BJ_SyncData:
list = &transaction->t_sync_datalist;
break;
case BJ_Metadata:
transaction->t_nr_buffers++;
list = &transaction->t_buffers;
break;
case BJ_Forget:
list = &transaction->t_forget;
break;
case BJ_IO:
list = &transaction->t_iobuf_list;
break;
case BJ_Shadow:
list = &transaction->t_shadow_list;
break;
case BJ_LogCtl:
list = &transaction->t_log_list;
break;
case BJ_Reserved:
list = &transaction->t_reserved_list;
break;
case BJ_Locked:
list = &transaction->t_locked_list;
break;
}
__blist_add_buffer(list, jh);
jh->b_jlist = jlist;
if (was_dirty)
set_buffer_jbddirty(bh);
}
void journal_file_buffer(struct journal_head *jh,
transaction_t *transaction, int jlist)
{
jbd_lock_bh_state(jh2bh(jh));
spin_lock(&transaction->t_journal->j_list_lock);
__journal_file_buffer(jh, transaction, jlist);
spin_unlock(&transaction->t_journal->j_list_lock);
jbd_unlock_bh_state(jh2bh(jh));
}
/*
* Remove a buffer from its current buffer list in preparation for
* dropping it from its current transaction entirely. If the buffer has
* already started to be used by a subsequent transaction, refile the
* buffer on that transaction's metadata list.
*
* Called under journal->j_list_lock
*
* Called under jbd_lock_bh_state(jh2bh(jh))
*/
void __journal_refile_buffer(struct journal_head *jh)
{
int was_dirty;
struct buffer_head *bh = jh2bh(jh);
J_ASSERT_JH(jh, jbd_is_locked_bh_state(bh));
if (jh->b_transaction)
assert_spin_locked(&jh->b_transaction->t_journal->j_list_lock);
/* If the buffer is now unused, just drop it. */
if (jh->b_next_transaction == NULL) {
__journal_unfile_buffer(jh);
return;
}
/*
* It has been modified by a later transaction: add it to the new
* transaction's metadata list.
*/
was_dirty = test_clear_buffer_jbddirty(bh);
__journal_temp_unlink_buffer(jh);
jh->b_transaction = jh->b_next_transaction;
jh->b_next_transaction = NULL;
[PATCH] jbd: fix BUG in journal_commit_transaction() Fix possible assertion failure in journal_commit_transaction() on jh->b_next_transaction == NULL (when we are processing BJ_Forget list and buffer is not jbddirty). !jbddirty buffers can be placed on BJ_Forget list for example by journal_forget() or by __dispose_buffer() - generally such buffer means that it has been freed by this transaction. Freed buffers should not be reallocated until the transaction has committed (that's why we have the assertion there) but they *can* be reallocated when the transaction has already been committed to disk and we are just processing the BJ_Forget list (as soon as we remove b_committed_data from the bitmap bh, ext3 will be able to reallocate buffers freed by the committing transaction). So we have to also count with the case that the buffer has been reallocated and b_next_transaction has been already set. And one more subtle point: it can happen that we manage to reallocate the buffer and also mark it jbddirty. Then we also add the freed buffer to the checkpoint list of the committing trasaction. But that should do no harm. Non-jbddirty buffers should be filed to BJ_Reserved and not BJ_Metadata list. It can actually happen that we refile such buffers during the commit phase when we reallocate in the running transaction blocks deleted in committing transaction (and that can happen if the committing transaction already wrote all the data and is just cleaning up BJ_Forget list). Signed-off-by: Jan Kara <jack@suse.cz> Acked-by: "Stephen C. Tweedie" <sct@redhat.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-06-23 09:05:25 +00:00
__journal_file_buffer(jh, jh->b_transaction,
jbd: fix possible journal overflow issues There are several cases where the running transaction can get buffers added to its BJ_Metadata list which it never dirtied, which makes its t_nr_buffers counter end up larger than its t_outstanding_credits counter. This will cause issues when starting new transactions as while we are logging buffers we decrement t_outstanding_buffers, so when t_outstanding_buffers goes negative, we will report that we need less space in the journal than we actually need, so transactions will be started even though there may not be enough room for them. In the worst case scenario (which admittedly is almost impossible to reproduce) this will result in the journal running out of space. The fix is to only refile buffers from the committing transaction to the running transactions BJ_Modified list when b_modified is set on that journal, which is the only way to be sure if the running transaction has modified that buffer. This patch also fixes an accounting error in journal_forget, it is possible that we can call journal_forget on a buffer without having modified it, only gotten write access to it, so instead of freeing a credit, we only do so if the buffer was modified. The assert will help catch if this problem occurs. Without these two patches I could hit this assert within minutes of running postmark, with them this issue no longer arises. Thank you, Signed-off-by: Josef Bacik <jbacik@redhat.com> Cc: <linux-ext4@vger.kernel.org> Acked-by: Jan Kara <jack@ucw.cz> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-04-28 09:16:12 +00:00
jh->b_modified ? BJ_Metadata : BJ_Reserved);
J_ASSERT_JH(jh, jh->b_transaction->t_state == T_RUNNING);
if (was_dirty)
set_buffer_jbddirty(bh);
}
/*
* For the unlocked version of this call, also make sure that any
* hanging journal_head is cleaned up if necessary.
*
* __journal_refile_buffer is usually called as part of a single locked
* operation on a buffer_head, in which the caller is probably going to
* be hooking the journal_head onto other lists. In that case it is up
* to the caller to remove the journal_head if necessary. For the
* unlocked journal_refile_buffer call, the caller isn't going to be
* doing anything else to the buffer so we need to do the cleanup
* ourselves to avoid a jh leak.
*
* *** The journal_head may be freed by this call! ***
*/
void journal_refile_buffer(journal_t *journal, struct journal_head *jh)
{
struct buffer_head *bh = jh2bh(jh);
jbd_lock_bh_state(bh);
spin_lock(&journal->j_list_lock);
__journal_refile_buffer(jh);
jbd_unlock_bh_state(bh);
journal_remove_journal_head(bh);
spin_unlock(&journal->j_list_lock);
__brelse(bh);
}